Measuring reactive nitrogen emissions from point sources using visible spectroscopy from aircraft

Abstract

Accurate measurements of nitrogen dioxide (NO₂), a key trace gas in the formation and destruction of tropospheric ozone, are important in studies of urban pollution. Nitrogen dioxide column abundances were measured during the Texas Air Quality Study 2000 using visible absorption spectroscopy from an aircraft. The method allows for quantification of the integrated total number of nitrogen dioxide molecules in the polluted atmosphere and is hence a useful tool for measuring plumes of this key trace gas. Further, we show how such remote-sensing observations can be used to obtain information on the fluxes of nitrogen dioxide into the atmosphere with unique flexibility in terms of aircraft altitude, and the height and extent of mixing of the boundary layer. Observations of nitrogen dioxide plumes downwind of power plants were used to estimate the flux of nitrogen oxide emitted from several power plants in the Houston and Dallas metropolitan areas and in North Carolina. Measurements taken over the city of Houston were also employed to infer the total flux from the city as a whole.

---

1. Introduction

Air pollution can harm the health of humans and wildlife, and can damage vegetation and natural water systems. To reduce the heath effects and environmental damage caused by air pollution, Congress passed and the United States Environmental Protection Agency (EPA) implemented the Clean Air Act in 1970. The Clean Air Act implemented National Ambient Air Quality Standards (NAAQS) to improve the quality of the air. The NAAQS are based on a few common air pollutants called criteria air pollutants. The criteria pollutants include tropospheric ozone (O₃), volatile organic compounds (VOCs), nitrogen dioxide (NO₂), carbon monoxide (CO), particulate matter, sulfur dioxide (SO₂), and lead (Pb). The abundance of the criteria pollutants must be less than the national control levels that are based on our understanding of human health effect to avoid a violation. Secondary standards designed to prevent environmental and property damage must also be met.

A criteria pollutant of primary interest is tropospheric ozone that is produced in a series of chemical reactions. The two main reactants are radicals produced from hydrocarbons (RO₂) and reactive nitrogen (NO_X) as identified by Haagen-Smit¹ in the 1950's. Reactive nitrogen is the sum of nitrogen dioxide (NO₂) and nitric oxide (NO). Once the nitrogen is in the form of NO₂, it photolyzes to give an oxygen atom that can rapidly react with O₂ to form O₃. This process is affected by the level of reactive hydrocarbons which oxidize NO to NO₂ thus resulting in the formation of more ozone.

A major source of NO_X is power plants which emit NO that in turn is readily oxidized to NO₂. In 1995 the US EPA² estimated that electric utilities are the largest single source of NO_X, emitting approximately 6233 short tons of NO_X per year.

In 1990 the Clean Air Act Amendments (CAAA) required stationary sources to install continuous emission monitors (CEMS) to measure the release of pollutants from their stacks. The CEMS provide a site-specific estimate of the emissions of NO_X from power plants. It has been argued that the CEMS tend to over estimate the emissions due to protocols that bias or introduce uncertainty, Placet et al.³ These protocols include the calibration of flow meters, missing data substitution protocols, data handling and software errors, and monitoring data that is located outside of the design range. Thus, alternative methods for checking NO_X emissions from power plants can benefit both industry, science, and regulators.

In a recent study, Ryerson et al.⁴ took in-situ measurements of NO₂ and NO using a chemiluminescence instrument from aircraft and calculated NO_X fluxes from several point sources. These measurements indicate the local level of NO_X in parts per billion by volume (ppbv) in the air mass the plane was flying through, so that to calculate total NO_X fluxes it was assumed that the boundary layer was completely well-mixed. The height of the boundary layer is also needed to obtain fluxes from in-situ data, along with the average wind speed.

Hoff and Millán⁵ suggested that visible absorption spectroscopy could be used to measure the fluxes of two major pollutants, SO₂ and NO_X. These pollutants can be measured by a remote sensing spectrograph because these gases exhibit strongly wave-length dependent absorption spectra in the near ultraviolet and visible regions. The measured absorption is related to the total integrated column abundance of each gas along the optical path (in molecules per square centimeter). Hence the absorption techniques provide important insights to the total burden of emitted molecules irrespective of where the molecules reside. Millán and others used a correlation spectrograph, COSPEC, to measure SO₂ from point sources, Hoff and Millán⁵ and reference therein.

Here we focus on NO₂ sources, and we compare and contrast in-situ information with column data. We also make use of simultaneous upward-looking and downward-looking instruments aboard the NCAR Electra research aircraft to further refine this method of flux measurement. We use miniature differential absorption spectroscopy (MIDAS) to take measurements of the total column abundances of NO₂ from the aircraft and estimate the amount of reactive nitrogen emitted from power plants. The remote total column measurement yields integrated NO₂ abundances in molecules per square centimeter. As we will show, these data can provide fluxes in molecules per second irrespective of the aircraft altitude and the height and uniformity of the boundary layer. Our work shows that visible spectroscopy is a very useful technique for measuring NO_X fluxes from point source and eliminates some previous errors discussed in estimating point source emissions.

The study focused on Houston, Texas, a city with numerous sources of NO_X. These sources are dominated by power plants and also include oil refineries, industrial plants, urban sources, and shipping. The study is referred to as the Texas Air Quality Study (TexAQS) and took place during the months of August and September, 2000. The complement of instruments aboard the Electra included the absorption spectrograph MIDAS as well as an in-situ chemiluminescence NO₂ and NO instrument. In this paper we will describe the MIDAS instrument and its technique. Information regarding the in-situ chemiluminescence instrument is given in Ryerson et al.,⁶ and references therein.

2. MIDAS instrumentation and technique

The core of MIDAS is two commercially purchased crossed Czerny-Turner fixed grating spectrographs. We cool the spectrographs to −20 °C using a commercial Marvel refrigerator designed for storage of medical vaccines at constant temperature to improve the stability of the dark current and reduce noise. A slit width of 50 µm in conjunction with a 1200 lines mm⁻¹ grating produces a full-width-maximum resolution of approximately 1.0 nm over the wavelength range from 425 to 725 nm. The incoming light is collected by downward and upward viewing lenses and directed into each spectrograph using fiber optics. The field of view is approximately 1° for the data taken from the Electra.

Our analysis employs the differential optical absorption spectroscopy (DOAS) technique that is extensively discussed in Sanders⁷ and references therein. The basis of the analysis is an equation similar in the form to the Beer-Lambert law that relates the optically thin attenuation of monochromatic radiation to the number of absorbing molecules in the optical path, Daniel et al.⁸ The Beer-Lambert law can be written as

where I(λ) and I₀(λ) are, according to spectroscopy terminology, the foreground and background spectra respectively, β_m is the differential slant column abundance of the mth species, and Q_m(λ) is the wavelength dependent absorption cross-section of the mth species. The background spectrum is measured in clear skies and is characterized by the total NO₂ in the non-polluted atmosphere (which resides mainly in the stratosphere). The foreground spectrum contains enhancements in NO₂ absorption in the atmosphere measured during the flights. The foreground and background spectra are taken with the same instrument which eliminates the need to characterize the instrument response function since the measurement is relative. The ratio between the foreground and the background measurements provides the difference in slant column abundance between the two, β_m.

The wavelength dependent absorption cross-sections, Q_m(λ), contains model functions that are linearly fit to the natural logarithm of the ratio of the foreground and background spectra (eqn. (1)). The model functions include the absorption cross-section of NO₂ and other gases such as ozone, Rayleigh scattering and representation of the Ring effect, Solomon et al.⁹ and Sanders.⁷ The absorption cross-section for NO₂ used in this analysis is based upon the study by Harder et al.¹⁰ A term proportional to the inverse wavelength is also included to approximate the cross-section for large-particle Mie scattering, i.e. to account for clouds or large aerosol particles in the optical path. The inversion procedure draws upon the unique spectral fingerprint of each molecule after removal of the smooth terms (i.e. the differential rather than absolute absorption). The slant column abundance, β_m, is multiplied by the absorption cross-section, Q_m(λ), to give the total column abundance of the NO₂ (Fig. 1), Sanders⁷ and Solomon et al.⁹

Fig. 1 Comparison of laboratory and TexAQS 2000 measurements of the differential absorption of radiation by nitrogen dioxide in the wavelength region 470–505 nm. This corresponds to a NO₂ column of 7.66 × 10¹⁶ molecules cm⁻² taken while flying through a plume downwind of the Monticello Power Plant in Texas.

3. Analysis

3.1 Reactive nitrogen flux calculation

Using MIDAS measurements, we calculate the flux of NO₂ from point sources using a modified form of the equations presented in White et al.,¹¹ Trainer et al.,¹² and Ryerson et al.⁴ MIDAS measures the total column abundance of NO₂ and therefore, the flux is assumed to be given by where v is the wind speed, α is the angle between the aircraft track and the direction normal to the wind direction, and the integral represents the area under the plumes over the time of the plume enhancement. Eqn. (2) assumes the wind speed and direction, plant emissions, and the flight speed are constant over the time period considered. These assumptions suggest that the measured NO₂ column integrated along the flight path will be constant if the measurements from multiple passes of the plume occurred the same distance downwind from the power plant. We test the validity of this assumption below.

Power plants report their estimated reactive nitrogen fluxes emitted in molecules per second. To compare our NO₂ vertical column measurements to the hourly fluxes reported by the power plants, we assume that photostationary state equilibrium is established between the NO and NO₂. We use the measured in-situ data for NO and NO₂ to determine the ratio of NO_X to NO₂. Multiplying our MIDAS NO₂ flux by this ratio results in NO_X emitted from the power plants according to MIDAS measurements (eqn. (3)).

3.2 Air mass factor determination

The MIDAS spectrographs are deployed in both the up-looking and down-looking directions, which allows the total amount of the enhanced NO₂ column along both lines of sight (slant columns) to be measured. In order to convert these slant column measurements to the vertical NO₂ column, the average path length of the photons detected (or air mass factor, AMF) must be known for the look directions and absorber profiles.

The air mass factor was determined using a radiative transfer model and checked using oxygen absorption features readily measured with MIDAS. Because the abundance of oxygen is well known (21% of the dry air composition) observations of its absorption provide a direct measure of the optical path or air mass factor; changes in oxygen absorption must reflect path differences, not changes in oxygen abundance. The observed air mass factor for oxygen is the measured amount of oxygen along the line of sight from MIDAS across the flight leg of interest divided by the amount of integrated oxygen in a vertical column at sea level.

We find that the oxygen AMF from observations agrees well with a radiative transfer model, providing confidence in the model. The model is a plane-parallel calculation based on the DISORT radiative transfer code, Stamnes et al.,¹³ which gives an accurate representation of multiple scattering effects, Portmann et al.¹⁴ The error in the AMF factor determined by comparison of the radiative transfer model compared to the measured AMF for oxygen ranges from 1–20%. Such differences probably stem from uncertainties in factors such as aerosol content, which can induce multiple scattering and hence the AMF.

If the vertical distribution of oxygen were the same as that of NO₂, we could rely upon our measurements of oxygen to provide air mass factors for direct use in the NO₂ analysis. However, air mass factors depend to some extent upon the vertical distribution of the absorber; the NO₂ is located largely in the boundary layer while oxygen is distributed throughout the atmosphere. To determine the AMF factor for NO₂ we used the radiative transfer model calculations. The error in the calculations is estimated to be the same error found between the measured and calculated AMF for oxygen. To first order, the AMF for measurements of NO₂ by MIDAS is approximately

where SZA is the solar zenith angle. The factor Γ depends on the surface albedo (A_s), the amount of NO₂ seen in the up-looking versus the down-looking instrument, and to a lesser extent on the altitude of the aircraft. The surface albedo used in the radiative transfer model is 0.2, the aircraft altitude was approximately one kilometer, and the absorber, NO₂, was assumed to be located in the range of 0–2 km. Note that Γ = 1 is the single scattering approximation if all photons either scatter above the absorbing layer or at the ground.

4. Results and discussion

We focus our analysis on two specific days: The flight of September 3rd, 2000 in which five passes were made through the plumes of the Welsh and Monticello power plants near Dallas at the same distance downwind and the flight of August 28th, 2000 in which two passes of the Houston area plume were made at different altitudes.

4.1 Multiple passes of plumes at same altitude

Fig. 2a shows the location of the two power plants and the flight leg where the five passes of September 3rd, 2000 occurred downwind from the Welsh and Monticello power plants. The flight path is shown as a function of NO₂ measurements from MIDAS in molecules per square centimeter.

Fig. 2 a–b Flight path on September 3rd, 2000 as a function of NO₂ molecules per square centimeter measured by MIDAS. b shows the MIDAS NO₂ measurement and the in-situ NO₂ measurement in ppbv. The black box indicates the flight leg of interest. Five passes at this distance downwind are analyzed.

The enhancements seen in the five passes of the flight leg downwind from the Monticello and Welsh power plants are also plotted as a time series in Fig. 2b. Included in this time series are the in-situ chemiluminescence NO₂ measurements in parts per billion by volume (ppbv). The in-situ and the MIDAS measurements show corresponding enhancements of NO₂ from the Monticello and Welsh power plants. The duration of each pass of the flight leg took approximately 25 min. The 1 h CEMS emissions data suggests plant emissions were relatively constant.

Fig. 2b shows that multiple passes of the same plant's plume are of similar magnitude in the MIDAS instrument. The plumes display notably more variability in the local in-situ measurement for these passes. This illustrates the difficulty in inferring total fluxes from in-situ data when constituents are not well-mixed; the MIDAS instrument is not subject to this limitation because the observations provide integrated abundances and are not dependent on the degree of mixing. Wind speed, direction, solar zenith angle, and the NO_X/NO₂ ratio are additional factors that enter into the calculation of the total flux. Averaged wind speed and direction were obtained from on-board aircraft measurements for the flight legs of interest in Fig. 2a. This average wind speed was 3.38 m s⁻¹ with a standard deviation of 0.99. Thus, the measured wind speed suggests up to a 30% uncertainty due to this variability. The wind direction was 337° out of the northwest. The solar zenith angle was 27° at the time the data was taken. The flight speed is the averaged flight speed across the length of the plume. The NO_X/NO₂ ratio was obtained from using the in-situ chemiluminescence measurements.

As discussed above the AMF was determined using the radiative transfer model. The resulting AMF was 1.83 for NO₂ in the boundary layer with a 10% error based upon the comparison of measured oxygen AMF and the calculated oxygen AMF. Fig. 3 shows the aircraft altitude in meters and measured oxygen column abundance by MIDAS in molecules per square centimeter. The oxygen AMF is shown for the measured and modeled calculations. The altitude and oxygen column abundances are consistent across the time series of the Monticello and Welsh power plant plumes providing support for the assumption of a nearly constant air mass factor on this flight leg.

Fig. 3 Aircraft altitude in meters and measured oxygen column by MIDAS in molecules per square centimeter during the time series of the flight leg through the Monticello and Welsh power plant plumes.

Measured net plume NO_X flux ranges from eqns. (2), (3), and (4) for the Welsh and Monticello power plants are shown in Table 1. Plumes W4 and M1 (Fig. 2b) were not used to calculate a flux because the aircraft turned before flying through the entire plume. Table 1 also shows the hourly reported NO_X emissions by the power plants 1 h CEMS average data. The largest uncertainties are due to the wind speed and the AMF. Errors in the 1 h CEMS average data are discussed in the Introduction.

Table 1 Summary of MIDAS NO_X fluxes for Monticello and Welsh power plants^a

Monitcello and Welsh power plant plumes
Plume	Plant's estimated CEMS NOX flux	MIDAS NOX flux
a Fluxes are given in units of 10^24 molecules s^−1. The plants estimated NOX fluxes are from 1 h CEMS averages.
Welsh 1	10.90	5.26 to 9.94
Welsh 2	10.90	5.39 to 10.18
Welsh 3	10.90	4.99 to 9.43
Monticello 2	9.76	5.18 to 9.77
Monticello 3	9.76	4.40 to 8.31
Monticello 4	9.76	4.86 to 9.19

---

As noted earlier, the reported NO_X fluxes from the power plant companies are believed to be accurate to within 30% and are more likely to be over estimated than underestimated using CEMS, Placet et al.³ and Ryerson et al.⁴ While the differences in NO_X fluxes from our measurements and the reported NO_X fluxes from the Monticello and Welsh power plants are within the combined uncertainties of the two, the data support the reported fluxes, as well as the view that they are probably overestimates rather than underestimates. The consistency of the measurements of NO₂ by MIDAS between multiple passes of the flight leg of interest provides important support for the analysis procedure.

4.2 Multiple passes over and through plumes at varying altitude

The second flight on which we focus took place on August 28th, 2000. During this flight two passes over the entire Houston, Texas area were made. Winds were from the southeast of the Gulf of Mexico, and the enhancements in the NO₂ as the air moved over the city could readily be identified, allowing quantification of the plume of the city as a whole. The first pass occurred at an aircraft altitude of 3500 m (above the boundary layer) and the second pass occurred at an aircraft altitude of 660 m (within the boundary layer). Fig. 4a and b show the time series for the two plumes.

Fig. 4 a–b Two passes of the Houston, Texas area NO₂ plume at two different altitudes. During the first plume pass the aircraft altitude was 3500 m and MIDAS sees NO₂ in the down-looking field of view only. During the second plume the aircraft altitude was 660 m and MIDAS sees NO₂ in both the down- and up-looking fields of view.

Fig. 4b shows the NO₂ concentration measured by MIDAS in molecules per square centimeter for the down- and up-looking fields of view separately. When the aircraft is flying at a high altitude MIDAS sees an enhancement of NO₂ in the down-looking field of view only. This is expected because we were not flying within the boundary layer; the enhancement in NO₂ from Houston should only be seen in the down-looking field of view. In contrast, the second Houston plume pass occurred at an aircraft altitude of 660 m. This pass occurred within the boundary layer and therefore an enhancement in NO₂ is seen in both the up-looking field of view and in the down-looking field of view. In-situ observations within the boundary layer showed that NO and NO₂ represented 77 ± 6% of the total reactive nitrogen averaged over this plume (with the remaining 23% largely being HNO₃); this value was used in converting NO₂ data to reactive nitrogen.

The results of the analysis using eqns. (2), (3), and (4) are shown in Table 2 for the two passes of the Houston area plume. The wind speed from the aircraft instrument was unavailable for this day. Therefore, the wind speed was assumed to be similar to August 27th, 2000. A sounding from the Lake Charles Station located at −93.21° longitude and 30.11° latitude on August 29th at 00Z suggests that this assumption is accurate to within ±1 ms⁻¹. This is an added uncertainty in the analysis of the Houston area plumes.

Table 2 Summary of MIDAS NO_X fluxes for Houston, Texas area^a

Aircraft altitude/m	Integrated Houston point source emissions	MIDAS NOX flux
a Fluxes are given in units of 10^24 molecules s^−1. The integrated Houston point source emissions is based on point sources from 28.8° latitude to 29.8° latitude. The point source emissions data is from EPA NET 96, version 3.12.^15
3500	142.00	122.15 to 212.24
660	142.00	106.74 to 184.36

---

The similarity of the results show that MIDAS measures the total column of NO₂ whether or not the aircraft is flying within the boundary layer. Further, as noted earlier, they also support the notion that the extent of mixing within the boundary layer and its height do not need to be known provided both up- and down-looking instruments are employed simultaneously. The measured NO_X flux by MIDAS agrees very well with the integrated running sum of point source emissions from the Houston area. The Houston area is defined as the point source in the range of 28.8 to 29.8° latitude and 96.1 to 94.8° west longitude.

4.3 Other plumes of interest

During TexAQS 2000, the aircraft passed through the W.A. Parish power plant in suburban Houston on several days. The analysis of the Parish power plant is presented in Table 3. On August 30th, 2000 two passes of the W.A. Parish power plant plume occurred at different altitudes. During the first pass, the aircraft altitude was 3500 m and during the second pass the aircraft altitude was 640 m. The larger error on the August 30th, 2000 Parish plumes is due to a greater variation in the wind speed than on other days analyzed. On August 28th, 2000 we flew through the W.A. Parish power plant plume at 656 m. The resulting NO_X flux on this day from the W.A. Parish power plant corresponds to approximately 5% of the Houston area plume analyzed above on the same day.

Table 3 Summary of MIDAS NO_X fluxes for Parish and North Carolina power plants^a

W.A. Parish station plumes
Date	Plant's estimated CEMS NOX flux	MIDAS NOX flux
8/27/00	10.88	4.00 to 7.91
8/28/00	14.57	5.79 to 10.37
8/30/00 3500m	15.17	7.66 to 22.83
8/30/00 640m	15.17	2.82 to 8.38

---

North Carolina 05/11/00
Plant	Plant's estimated CEMS NOX flux	MIDAS NOX flux
a Flux are given in units of 10^24 molecules s^−1. The plants estimated NOX fluxes are from 1 h CEMS averages.
Roxboro	11.6	8.56 to 16.48
Belews	10.7	4.54 to 8.73

---

Also presented in Table 3 are the NO₂ observations and estimated fluxes from another MIDAS flight on May 11th, 2000 in North Carolina through the Roxboro and Belews Creek power plant plumes. On this flight, no in-situ observations were taken, and we have assumed a ratio of NO_X/NO₂ as observed in Texas. The wind speed is also an estimated wind speed due to a failure in the aircraft instrument. The NO_X/NO₂ ratio and the wind speed estimates are added uncertainty factors for this flight. The measured NO_X fluxes from MIDAS and the reported NO_X fluxes from the Roxboro and Belews Creek power plants are within the combined uncertainties of the two fluxes.

5. Conclusions

Nitrogen dioxide slant column measurements taken using visible absorption spectroscopy from aircraft in the Houston and Dallas, Texas areas and in North Carolina were used to calculate NO_X fluxes from power plants and for the city of Houston. The reported 1 h average CEMS NO₂ estimated emissions data from the Monticello and Welsh power plants agree within error estimates with the measured MIDAS NO_X flux. However, in most (but not all) cases, the reported 1 h average CEMS NO₂ estimated emissions data appear likely to be biased high compared to the measured MIDAS NO_X flux. The Parish power plant flux as measured by MIDAS is generally smaller than the plant estimate with one exception. For the two plants in North Carolina, the Belews Creek power plant suggests the 1 h CEMS data is an overestimate whereas the Roxboro plant 1 h CEMS data agrees well with the measured MIDAS flux.

The analysis presented is consistent across multiple passes of the plumes at the same altitude and same distance downwind from the power plants as shown with the Monticello and Welsh power plants. The consistency of the measurements of NO₂ by MIDAS between multiple passes of the flight leg of interest support the assumption of constant wind speed, wind direction, flight speed, and plant emissions. The MIDAS method is independent of the aircraft altitude and uniformity of the boundary layer as shown in the analysis of the Houston plume on August 28th, 2000 and the Parish plume on August 30th, 2000. Thus, our NO₂ measurements using MIDAS complement in-situ methods and can allow for a more comprehensive understanding of point source emissions.

References

1. A. J. Haagen-Smit, Ind. Eng. Chem., 1952, 44, 1342.
2. Environmental Protection Agency (EPA), 1996, National Air Pollution Emission Trends, 1900–1995, EPA-454/R-969007, Office of Air Quality Planning and Standards, Research Triangle Park, NC, 2000.
3. M. Placet, C. O Mann, R. O Gilbert and M. J. Niefer, Atmos. Environ., 2000, 34, 2183.
4. T. B. Ryerson, M. P. Buhr, G. J. Frost, P. D. Goldan, J. S. Holloway, G. Hübler, B. T. Jobson, W. C. Kuster, S. A. McKeen, D. D. Parrish, J. M. Roberts, D. T. Sueper, M. Trainer, J. Williams and F. C. Fehsenfeld, J. Geophys. Res., 1998, 103, 22–569.
5. R. M. Hoff and M. M. Millán, Air Pollut. Control Assoc., J., 1981, 31(4), 381.
6. T. B. Ryerson, E. J. Williams and F. C. Fehsenfeld, J. Geophys. Res., 2000, 105, 26–447.
7. R. W. Sanders, J. Geophys. Res., 1996, 101, 20–945.
8. J. S. Daniel, S. Solomon, R. W. Portmann, A. O. Langford, E. G. Dutton and W. Madsen, J. Geophys. Res., 2002.
9. S. Solomon, G. H. Mount, R. W. Sanders and A. L. Schmeltekopf, J. Geophys. Res., 1987, 92, 8329.
10. J. W. Harder, J. W. Brault, P. V. Johnston and G. H. Mount, J. Geophys. Res., 1997, 102, 3861.
11. W. H. White, J. A. Anderson, D. L. Blumenthal, R. B. Husar, N. V. Gillani, J. D. Husar and W. E. Wilson, Jr, Science, 1976, 194, 187.
12. M. Trainer, B. A. Ridley, M. P. Buhr, G. Kok, J. Walega, G. Hübler, D. D. Parrish and F. C. Fehsenfeld, J. Geophys. Res., 1995, 100, 18–823.
13. K. Stamnes, S. C. Tsay, W. Wiscombe and K. Jayaweera, Appl. Opt., 1988, 27, 2502.
14. R.W. Portmann, S. Solomon, R. W. Sanders and J. S. Daniel, J. Geophys. Res., 2001, 106, 1139.
15. Environmental Protection Agency (EPA), 2000, National Air Pollutant Emission Trends, 1990–1998, EPA-454/R-00-002, Office of Air Quality Planning and Standards, Research Triangle Park, NC, 2000.

---

Footnote

† Also at Cooperative Institute for Research in Environmental Sciences, University of Colorado, Boulder, Colorado, USA.

---